1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
use crate::prelude::*;
use super::*;
use reqwest::Client;
use futures_util::StreamExt;
use tokio::sync::mpsc;
use tokio_stream::wrappers::UnboundedReceiverStream;
/// The LM Studio chat
pub struct Chat {
pub(crate) model: Model,
pub(crate) context: Context,
pub(crate) host: String,
pub(crate) client: Client,
pub(crate) reader: Option<ResponseReader>,
}
impl Chat {
/// Creates a new simple chat
pub fn new<M: Into<Model>, C: Into<Context>>(model: M, context: C, port: u16) -> Self {
let chat = Self {
model: model.into(),
context: context.into(),
host: fmt!("http://localhost:{port}"),
client: Client::new(),
reader: None,
};
/*
// loading model:
chat.load_model(chat.model.clone()).await?;
*/
chat
}
/*
/// Loads AI model to memory
pub async fn load_model<M: Into<Model>>(&mut self, model: M) -> Result<()> {
let request = Prompt {
model: model.into(),
prompt: str!("Hello"),
stream: false,
..Default::default()
}.into();
let _ = self.send(request).await?;
Ok(())
}
*/
/// Send request to chat
pub async fn send(&mut self, request: Request) -> Result<Option<Response>> {
self.context.update_system_info().await;
match request {
Request::Messages(request) => self.handle_messages(request).await,
Request::Prompt(request) => self.handle_prompt(request).await,
Request::Embeddings(request) => self.handle_embeddings(request).await,
}
}
/// Handle messages request
async fn handle_messages(&mut self, mut request: Messages) -> Result<Option<Response>> {
let url = fmt!("{}/v1/chat/completions", self.host);
// choose AI model:
if let Model::Other(s) = &request.model {
if s.is_empty() {
request.model = self.model.clone();
}
}
// add request to context:
request.messages = if request.context {
for msg in request.messages {
self.context.add(msg);
}
self.context.get()
} else {
let mut context = self.context.clone();
for msg in request.messages {
context.add(msg);
}
context.get()
};
// handle request:
if !request.stream {
let response = self.client.post(&url)
.json(&request)
.send()
.await?
.error_for_status()?
.json::<Response>()
.await?;
/*
// filtering <think>..</think> block:
if request.skip_think {
let re = re!(r"(?s)<think>.*?</think>");
for choice in &mut response.choices {
let message = choice.message.as_mut().unwrap();
message.content = re.replace_all(&message.content, "").trim().to_string();
}
}
*/
// add response to context:
if request.context {
if let Some(choice) = response.choices.get(0) {
let message = choice.message.as_ref().unwrap();
let answer = Message::new(Role::Assistant, message.text());
self.context.add(answer);
}
}
Ok(Some(response))
} else {
// spawning stream reader:
self.spawn_reader(url.clone(), request.clone(), request.context, request.skip_think).await?;
Ok(None)
}
}
/// Handle prompt request
async fn handle_prompt(&mut self, mut request: Prompt) -> Result<Option<Response>> {
let url = fmt!("{}/v1/completions", self.host);
// choose AI model:
if let Model::Other(s) = &request.model {
if s.is_empty() {
request.model = self.model.clone();
}
}
// add request to context:
request.prompt = if request.context {
self.context.add(request.prompt);
self.context.get_as_string()
} else {
let mut context = self.context.clone();
context.add(request.prompt);
context.get_as_string()
};
// handle request:
if !request.stream {
let mut response = self.client.post(&url)
.json(&request)
.send()
.await?
.error_for_status()?
.json::<Response>()
.await?;
if request.skip_think {
let re = re!(r"(?s)<think>.*?</think>");
for choice in &mut response.choices {
let text = choice.text.as_mut().unwrap();
*text = re.replace_all(&text, "").trim().to_string();
}
}
Ok(Some(response))
} else {
// spawning stream reader:
self.spawn_reader(url.clone(), request.clone(), false, request.skip_think).await?;
Ok(None)
}
}
/// Handle embeddings request
async fn handle_embeddings(&mut self, mut request: Embeddings) -> Result<Option<Response>> {
let url = fmt!("{}/v1/embeddings", self.host);
// choose AI model:
if let Model::Other(s) = &request.model {
if s.is_empty() {
request.model = self.model.clone();
}
}
// handle request:
let response = self.client.post(&url)
.json(&request)
.send()
.await?
.error_for_status()?
.json::<Response>()
.await?;
Ok(Some(response))
}
/// Spawns stream reader
async fn spawn_reader<J>(&mut self, url: String, request: J, context: bool, skip_think: bool) -> Result<()>
where J: Serialize + Send + Sync + 'static,
{
let (tx, rx) = mpsc::unbounded_channel::<Result<StreamChoice>>();
let client = self.client.clone();
self.reader = Some( ResponseReader::new(UnboundedReceiverStream::new(rx), context) );
tokio::spawn(async move {
let mut is_thinking = false;
let mut is_after_thinking = false;
let response = client.post(&url)
.json(&request)
.send()
.await;
match response {
Ok(response) => {
let mut stream = response.bytes_stream();
while let Some(item) = stream.next().await {
match item {
Ok(chunk) => {
let chunk = String::from_utf8_lossy(&chunk);
for line in chunk.lines() {
// parsing response line:
if line.starts_with("data: ") {
let data = &line[6..];
if data == "[DONE]" {
break;
}
let stream: Result<Stream> = json::from_str(data).map_err(Into::into);
let stream = if let Ok(r) = stream { r }else{ continue };
for mut choice in stream.choices {
if let Some(text) = choice.text_mut() {
// filtering <think>..</think> block:
if skip_think {
if is_thinking {
if text.contains("</think>") {
is_thinking = false;
is_after_thinking = true;
}
continue;
}
else if text.contains("<think>") {
is_thinking = true;
continue;
}
// trim extra spaces:
if is_after_thinking {
*text = text.trim_start().to_string();
is_after_thinking = false;
}
}
// send answer part to channel:
if tx.send(Ok(choice)).is_err() {
break;
}
} else {
continue;
}
}
}
}
},
Err(e) => {
let _ = tx.send(Err(e.into()));
break;
}
}
}
},
Err(e) => {
let _ = tx.send(Err(e.into()));
}
}
});
Ok(())
}
/// Read next stream choice
pub async fn next(&mut self) -> Option<Result<StreamChoice>> {
if let Some(reader) = &mut self.reader {
let result = reader.next().await;
if reader.context && reader.is_ready {
self.context.add(reader.message.clone())
}
result
} else {
self.reader = None;
None
}
}
}